Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,49 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
# ContentV: Efficient Training of Video Generation Models with Limited Compute
|
| 6 |
+
|
| 7 |
+
This project presents ContentV, a novel framework that accelerates DiT-based video generation through three key innovations:
|
| 8 |
+
- A minimalist model design that enables effective reuse of pre-trained image generation models for video synthesis
|
| 9 |
+
- A comprehensive exploration of a multi-stage, efficient training strategy based on Flow Matching
|
| 10 |
+
- A low-cost Reinforcement Learning with Human Feedback (RLHF) approach that further enhances generation quality without the need for additional human annotations.
|
| 11 |
+
|
| 12 |
+
## Quickstart
|
| 13 |
+
|
| 14 |
+
#### Recommended PyTorch Version
|
| 15 |
+
|
| 16 |
+
- GPU: torch >= 2.3.1 (CUDA >= 12.2)
|
| 17 |
+
- NPU: torch and torch-npu >= 2.1.0 (CANN >= 8.0.RC2). Please refer to [Ascend Extension for PyTorch](https://gitee.com/ascend/pytorch) for the installation of torch-npu.
|
| 18 |
+
|
| 19 |
+
#### Installation
|
| 20 |
+
|
| 21 |
+
```sh
|
| 22 |
+
git clone https://github.com/bytedance/ContentV.git
|
| 23 |
+
pip3 install -r ContentV/requirements.txt
|
| 24 |
+
```
|
| 25 |
+
|
| 26 |
+
#### T2V Generation
|
| 27 |
+
|
| 28 |
+
```sh
|
| 29 |
+
cd ContentV
|
| 30 |
+
## For GPU
|
| 31 |
+
python3 demo.py
|
| 32 |
+
## For NPU
|
| 33 |
+
USE_ASCEND_NPU=1 python3 demo.py
|
| 34 |
+
```
|
| 35 |
+
|
| 36 |
+
## Todo List
|
| 37 |
+
- [x] Inference code and checkpoints
|
| 38 |
+
- [ ] Training code of RLHF
|
| 39 |
+
|
| 40 |
+
## License
|
| 41 |
+
This code repository and part of the model weights are licensed under the [Apache 2.0 License](https://www.apache.org/licenses/LICENSE-2.0). Please note that:
|
| 42 |
+
- MM DiT are derived from [Stable Diffusion 3.5 Large](https://huggingface.co/stabilityai/stable-diffusion-3.5-large) and trained with video samples. This Stability AI Model is licensed under the [Stability AI Community License](https://stability.ai/community-license-agreement), Copyright © Stability AI Ltd. All Rights Reserved
|
| 43 |
+
- Video VAE from [Wan2.1](https://huggingface.co/Wan-AI/Wan2.1-T2V-14B) is licensed under [Apache 2.0 License](https://huggingface.co/Wan-AI/Wan2.1-T2V-14B/blob/main/LICENSE.txt)
|
| 44 |
+
|
| 45 |
+
## Acknowledgement
|
| 46 |
+
* [Stable Diffusion 3.5 Large](https://huggingface.co/stabilityai/stable-diffusion-3.5-large)
|
| 47 |
+
* [Wan2.1](https://github.com/Wan-Video/Wan2.1)
|
| 48 |
+
* [Diffusers](https://github.com/huggingface/diffusers)
|
| 49 |
+
* [HuggingFace](https://huggingface.co)
|